Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/crude-code/mcp-app/well-forecastingnpx skills add crude-code/mcp-app --skill well-forecastinggit clone --depth 1 https://github.com/crude-code/mcp-appWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00057 | $0.07949 |
| Opus 5 | $0.00028 | $0.03974 |
| Sonnet 5 | $0.00011 | $0.01590 |
| Haiku 4.5 | $0.00006 | $0.00795 |
Grade A, and why
well-forecasting scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 631 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Well Forecasting
The job
You are the reservoir engineer on this deal. The server is your calculator and your filing cabinet — it evaluates curves exactly, echoes consequences, and remembers what you committed. It never chooses a parameter. Every judgment is yours.
The objective is to predict future production. Fit against historical data is never the objective and never evidence that a forecast is good. History is evidence to be weighed, not a target to be matched.
Optimize the next 12 months. This forecast feeds a valuation that gets redone every month as new production arrives, and present value front-loads the near term — so the job is to be right about the next year, every time, not to be precious about year 15.
Where evidence is thin, borrow from the population or carry a range. Never fake precision.
Forecasting is not two methods — a decline-curve branch for producers and a type-curve branch for everything else. It is one method on a continuum: every forecast blends what the well has said with what its population says, weighted by how much the well has said. One reported month — you don't even know it's the peak — is forecast almost entirely from analogs. Seven months in, the level is the well's own but Di and b still come from the population, checked against what little slope the well shows. Three clean years and the well carries its forecast. The analog method (below) is not the non-producer branch; it is the population end of every forecast.
Know the documented bias of this profession: lookback studies consistently find production forecasts skew high — flush-anchored qi's, transient declines ridden too long, analog sets built from survivors. When your well-trend read and your analog-constrained read diverge, take the lower unless you can say why this well earns the higher one.
This is not pattern recognition. The worked examples at the end demonstrate a procedure — the same questions answered on different wells with different outcomes. They are not templates. Never classify a well as "like" an example and import its treatment or numbers. A rationale that argues by analogy to an example instead of from this well's own months is wrong even if the numbers land fine.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 631 lines · 57 tokens per session scan A 433e445c3030
well-forecasting is a skill published in the GitHub repository crude-code/mcp-app (4 stars, last pushed 3d ago), licensed Apache-2.0. It adds 57 tokens to every session and 7,949 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
monte-carlo
Use when the user wants Monte Carlo simulation of a PlanExe model — sampling from bounds to produce output distributions (mean/std/percentiles), threshold pass probabilities, and Pearson-correlation sensitivity rankings — given an extract-parameters-from-full JSON, a generate-bounds JSON, a generate-calculations…
fin-paper-figure
Generate academic-quality figures (>=300 DPI) for economics and finance papers.
de-portfolio-opt
差分进化（DE）组合优化。用户说「DE组合优化」「差分进化权重」「/de-portfolio-opt」时使用。长仅权重最大化样本夏普；纯 Python。默认 createweb。.
hht-timing
改进 HHT 择时（相位代理）。用户说「HHT」「希尔伯特」「EMD择时」「/hht-timing」时使用。无 EMD 库时用包络/相位近似并 meta.degraded。默认 createweb。.
quant-experiment-runtime
Quant research experiment executor: discover an offline source database under the workdir's code-repo, build a panel, run a Research Artifact's entry point to compute research-object values, and evaluate IC/ICIR/RANKIC/coverage metrics. Runtime = Experiment Executor; it runs a Research Artifact via a Python-native…
fin-paper-writing
经济金融论文写作编排器。根据PAPEROUTLINE.md和REFINEDDESIGN.md，编排调用fin-paper-draft（正文写作）、fin-paper-figure（图表生成）、fin-review-loop（review循环），管理版本并确保章节间的一致性。.